Improved Transient Response in Inverter-Based Resources using Deep Reinforcement Learning

Abstract

In this work, a deep reinforcement learning (DRL)-based controller is presented to improve the performance of an existing control process in inverter-based resources (IBR). A DRL algorithm is utilized to learn a model-free value function which further improves the transient response of an IBR over fast response time frames. In addition, the developed controller provides black-box control, that is, it performs the desired control action without accessing or modifying the internal parameters of the existing controller in an IBR. The developed controller is tested in case studies utilizing EMT simulations in PSCAD/EMTDC and its performance, resilience, and adaptiveness are verified.